Mercator Ocean Quarterly Newsletter
#26– July 2007 – Page 45 Equatorial wave intra-seasonal variability in the Indian and Pacific Oceans
Equatorial wave intra-seasonal variability in the Indian and Pacific Oceans in the Mercator-Ocean POG05B simulation By Serena Illig1, Boris Dewitte2, Claire Périgaud1 and Corinne Derval3 1
Jet Propulsion Laboratory (JPL/CALTECH), Pasadena, CA, USA LEGOS/IRD/IMARPE, Esquina de Gamarra y General Valle S/N Chucuito, Callao, Peru 3 Cerfacs in Mercator Ocean - 8-10 rue Hermes. 31520 Ramonville St-Agne 2
Introduction The tropical oceans have long been recognized as the most important region for large scale ocean-atmosphere interactions that give rise to coupled climate variations on several time scales. Whereas the tropical Atlantic and Pacific variability is the siege of air-sea interactions which characteristics are controlled to a large extend by equatorial wave dynamics (see Latif et al., 1998) for the Pacific and Xie and Carton (2004) for a recent review for the tropical Atlantic), the tropical Indian ocean is rather subject to the strong forcing by the Indian Ocean monsoon that has a strong meridional component. This makes the equatorial wave response to the zonal wind stress more difficult to identify than in the other oceans. In the three oceans, equatorial waves were detected from observations (Périgaud and Dewitte, 1996; Boulanger and Fu, 1996; Handoh and Bigg, 2000; LeBlanc and Boulanger, 2001) and model simulations (Dewitte et al., 2003; Illig et al., 2004; Yuan and Han, 2006) at seasonal to inter-annual timescales. To estimate the Kelvin and Rossby wave amplitude from altimetric data, the ‘one-mode’ approximation which assumes that the surface data are representative of a single baroclinic mode, is required because the subsurface is insufficiently sampled to derive the vertical mode variability. Although there are limitations linked to the use of non-biases-free model and assimilation scheme, Dewitte et al. (2003) showed for the tropical Pacific that products with data assimilation can bring further insight for the interpretation of the altimetric data and surface variability associated to the equatorial waves. For the Indian and Atlantic Ocean, the ‘one-mode’ approximation is anyway not valid due to the shallower thermocline and the peculiarities of the wind forcing. For instance, Illig et al. (2004) showed that 6 baroclinic modes are required at least to account for ~80% of the sea level variability. In this study, we take advantage of the Mercator-Ocean effort to design a global system with data assimilation using the ORCA model (Madec et al., 1998) to document the equatorial Indian Ocean vertical structure variability. Our study is based on the ‘free’ run (i.e. without data assimilation), which serves as a benchmark in order to assess the impact of data assimilation in future works. The paper identifies clear equatorial waves in the Indian Ocean and compares their characteristics to the ones in the Pacific Ocean focusing on the intra-seasonal variability.
Model simulation’description and validation In this study, we used a free simulation for the period 1992-2001 (Derval et al., 2005) of the Mercator-Ocean global model (see Dewitte et al., this issue, for a complete description of the simulation used). The simulation is referred to as POG05B hereafter. Because equatorial wave depends on the mean stratification, the mean temperature and salinity field are first compared to observations from the World Ocean Atlas 2001 (Conkright et al., 2002) (Figure 1). Figure 1 reveals known biases of many OGCMs, namely a too diffuse thermocline and a too salty warm pool. Overall, the simulated mean structure is rather realistic.
a) MERCATOR POG05B
Mercator Ocean Quarterly Newsletter
#26– July 2007 – Page 46 Equatorial wave intra-seasonal variability in the Indian and Pacific Oceans
b) Observations
Figure 1 Equatorial mean structure within the first 250 meters over the tropical Indian-Pacific sector (1993-2001) for a) MERCATOR POG05B simulation and for b) the observations (World Ocean Atlas 2001). Grey shading represents the mean temperature section between 16°C and 30°C. 34, 34.25, 34.5, 34.75, 35 iso-salinity are represented with red, green, light blue, blue and purple solid lines respectively. The model surface variability is then compared to satellite-derived observations, the TOPEX/Poseidon and ERS-1/2 combined data sets from October 1992 to January 2002 (Ducet et al., 2000) and the OSCAR current anomaly. OSCAR provides nearsurface currents derived from satellite altimeter, scatterometer and SST (Bonjean and Lagerloef, 2002). The surface layer current is the sum of geostrophic and Ekman currents and of a buoyancy term. The comparison indicates that the model simulate fairly well the surface variability in the equatorial band, with on average more skill in the Pacific than in the Indian oceans (Figure 2a) due to the energetic inter-annual variability in the Pacific (representative of ENSO) which is more easily grasped by the model than in the Indian ocean. Similar analysis on the high-pass filtered outputs (figure 2b) reveals however that the model simulate fairly well the intra-seasonal variability in the Indian ocean with correlation level as large as in the Pacific ocean for sea level and larger by ~0.1 on average than in the Pacific ocean for the surface currents.
a) Sea Level Anomalies
a) Surface Zonal Current Anomalies
Figure 2a Comparison between POG05B and independent satellite observations over the tropical Indian-Pacific sector over the 19932001 period. Sea Level (top panel) and Surface zonal current (bottom panel) inter-annual anomalies 99% significant correlations (red-shaded) as well as RMS differences between model and AVISO merged sea level (OSCAR surface currents) (color contours) are shown. Contour intervals are 2 cm for Sea Level RMS and 5 cm/s.for zonal current RMS.
Mercator Ocean Quarterly Newsletter
#26– July 2007 – Page 47 Equatorial wave intra-seasonal variability in the Indian and Pacific Oceans
a) Sea Level Anomalies
a) Surface Zonal Current Anomalies
Figure 2b -1
Same as Figure 2a, but for the high-pass filtered data (fc= (150) days ).
Vertical structure variability and wave sequences A vertical mode decomposition of the model mean stratification is sought following Dewitte et al. (1999) and baroclinic mode contribution to pressure and current anomalies are derived, which then allows inferring the Kelvin and Rossby wave amplitudes. Such method was used in previous studies for the equatorial Pacific (Dewitte el al., 1999; 2003) and Atlantic (Illig et al. 2004). It is extended here for the case of the Indian Ocean.
Baroclinic mode contributions We first analyze the characteristics of the baroclinic modes variability in terms of surface zonal current. The total Zonal Current Anomalies (ZCA) are estimated by averaging the currents over the 4 uppermost levels of the model (5–30 m depth), in order to remove the shear within the weakly stratified surface layer associated with incomplete mixing which cannot be represented in the vertical mode decomposition. The RMS variability of the surface ZCA is presented in Figure 3a.
Mercator Ocean Quarterly Newsletter
#26– July 2007 – Page 48 Equatorial wave intra-seasonal variability in the Indian and Pacific Oceans
Figure 3 Maps of variability (RMS) over 1993-2001 of (a) MERCATOR POG05B total surface zonal currents, (b-d) the contribution of the three first baroclinic modes, (e) the summed-up contribution of the high order baroclinic modes (4-8), and (f) the sum of the 8 gravest baroclinic modes. Contour interval is 5 cm/s. Values smaller than 6 cm/s appear in white; -1
Large values (>50 cm.s ) are confined within 2°S–2°N, with maximum variability centered around 75°W and in the western boundary current system. The results of the vertical mode decomposition of the zonal current variability (Figures 3b-f) indicate that the current variability project over a large number of modes, with the second baroclinic mode being the most energetic in the central basin. Interestingly, the wind projection coefficient along the equator is 0.73, 0.56, 0.18 and 0.32 for the first, second, third and fourth baroclinic mode on average in the western basin indicating that the mean stratification should favour the first baroclinic mode. This suggests a specific pattern of the wind forcing which amplify the second baroclinic mode contribution or -1 the presence of resonant modes. The second baroclinic mode is indeed expected to resonate at the 60, 90 and 180 days frequencies, timescales that are present in the wind forcing over the Indian Ocean (Shinoda et al., 1998). The explained variances of the mode contribution to sea level anomalies are on average along the equator 30%, 15%, 5% and 60% for respectively the first, second, third and the summed-up contributions of modes 1 to 8 for both the total and high-passed -1 filtered (fc=150 days ) variability.
Wave sequences Projecting the baroclinic mode contributions to zonal current and sea level anomalies onto the theoretical Kelvin and Rossby wave structures provide an estimation of the equatorial wave amplitude (Illig et al., 2004). The figures 4 and 5 display the result of the decomposition (second baroclinic mode) for both the total and high-pass filtered outputs. Clear propagations of the sea level anomaly can be observed.
Mercator Ocean Quarterly Newsletter
#26– July 2007 – Page 49 Equatorial wave intra-seasonal variability in the Indian and Pacific Oceans
a) Observations
c) Kelvin
b) POG05B
d) Rossby 1
e) Kelvin
Figure 4: Longitude-time plot of the seal level anomalies for a) the altimetric data (at 0°N), for b) the POG05B simulation (at 0°N), for c/e) the Kelvin wave contribution (K at 0°N) and the first meridional Rossby component (R1 at 3°S) for the second baroclinic mode. R1 is displayed reverse from 99°E to 43°E and K is repeated in order to visualize the reflection at the eastern and western basin boundaries. Positive (negative) values are red (blue) shaded respectively.
Mercator Ocean Quarterly Newsletter
#26– July 2007 – Page 50 Equatorial wave intra-seasonal variability in the Indian and Pacific Oceans
They are associated to the main sea level anomalies during this period, in particular during the 1997/98 El Niño event (figure 4). Whereas the 1997/98 El Niño in the Pacific is associated to the forcing of energetic downwelling Kelvin waves (Dewitte et al., 2003), in the Indian equatorial Ocean, upwelling Kelvin are forced in 1997 (figure 4c) which reflects at the eastern boundary as Rossby waves (figure 4d). POG05B also simulate propagations of equatorial waves at intra-seasonal frequencies (figures 5cde).
a) Observations
c) Kelvin
b) POG05B
d) Rossby 1
e) Kelvin
Figure 5 -1
Same as figure 4 but for the high-pass filtered anomalies (fc= (150) days ).
Intra-seasonal variability and role of boundary reflections Wavenumber-frequency diagrams In order to analyze the propagating nature of the estimated Kelvin and Rossby components, a bivariate space-time spectral analysis (Hayashi, 1977) is applied on the high-pass filtered POG05B outputs along the equator (for the Kelvin wave) and along st
3°S (for the 1 meridional Rossby wave component R1 wave) for the first two energetic baroclinic modes. Results are displayed in figures 6 and 7 for the Indian and Pacific Oceans respectively. The diagrams reveal the presence of long-wave length (k
